Analysis system for tobacco fermentation microbial community structure

Abstract

The invention discloses an analysis system for a tobacco fermentation microbial community structure, and belongs to the technical field of data processing, and the system comprises a sampling unit which is used for obtaining a tobacco leaf surface microbial sample so as to obtain original sequencing data through a high-throughput sequencing platform, and carrying out the data preprocessing to obtain a sequence file; the clustering unit is used for processing the preprocessed sequence by adopting a distance-based hierarchical clustering method, and determining a target clustering result through a nonlinear distance updating strategy and a clustering evaluation index; and the community structure analysis unit is used for constructing a phylogenetic classification tree based on the target clustering result, and obtaining fermentation microbial community structure characteristics of the tobacco samples in combination with diversity indexes among the samples. By means of the scheme, the clustering deviation can be reduced, and the influence of sequencing quality fluctuation on the clustering result is reduced.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to an analysis system for the structure of tobacco fermentation microbial communities. Background Art

[0002] In the existing technical field of tobacco fermentation microbial community structure analysis, sequence clustering analysis faces several key technical bottlenecks. The primary problem is that the existing technology generally adopts a sequence similarity calculation method based on edit distance, which uses a unified weight coefficient to process all types of sequence variations. This simple processing method has obvious technical defects in analyzing tobacco fermentation microbial sequences: when processing highly similar sequences with differences in key sites, it is impossible to accurately reflect the actual impact of these differences on sequence function; when calculating the similarity of sequences of different lengths, the lack of an effective standardization mechanism leads to a lack of comparability of distance values; at the same time, this method cannot adapt to differences in sequence quality caused by fluctuations in environmental factors during sequencing. These technical problems directly affect the quality of subsequent clustering analysis, resulting in strains with similar functions but slight differences in sequences being misclassified, or strains with significant functional differences but high sequence similarity being incorrectly aggregated. At the level of sequence clustering algorithms, the existing technology mainly relies on hierarchical clustering methods with fixed parameters. This type of method uses a simple linear distance update strategy, which shows serious technical limitations when dealing with tobacco microbial sequences with complex evolutionary relationships: the clustering process is easily affected by the initial parameter settings and falls into local optimality; the distance update rule lacks adaptability and cannot dynamically adjust parameters according to the sequence characteristics of different clustering stages; when processing large-scale sequence data, the fixed update strategy leads to low computational efficiency. These problems manifest themselves in practical applications as technical defects such as unstable clustering results, high consumption of computing resources, and insufficient clustering accuracy. Especially when analyzing the dynamic changes of microbial communities during fermentation, these technical defects will lead to the inability to accurately capture the succession characteristics of the community structure, affecting the monitoring and regulation of the fermentation process.

[0003] In addition, the existing technology also has obvious deficiencies in the evaluation and optimization of clustering results. Traditional methods usually use a fixed threshold to determine the number of clusters. This simple decision-making mechanism cannot adapt to sequence data of different quality levels. When selecting representative sequences, the existing technology often only considers the frequency of sequence occurrence, ignoring the impact of sequence quality on classification accuracy. This rough processing method easily leads to the lack of biological representativeness of the selected representative sequences, which in turn affects the accuracy of subsequent species annotation and function prediction. At the same time, the existing technology lacks an effective clustering quality evaluation mechanism and cannot objectively measure the reliability of clustering results, which makes it difficult to ensure the scientificity and repeatability of clustering analysis results. In the specific data processing link, the existing technology is insufficient in dealing with abnormal data and noise in the fermentation process. Due to the lack of an effective data quality control mechanism, the systematic errors and random noise generated during the sequencing process are easily transmitted to subsequent analysis, affecting the reliability of clustering results. At the same time, the existing methods lack flexible response strategies when dealing with abnormal situations such as sequence deletions and mismatches, which leads to problems such as data processing interruption or result deviation in practical applications. These technical defects seriously restrict the actual application effect of microbial community structure analysis methods in tobacco fermentation process control.

[0004] Therefore, there is an urgent need for a technical solution that can reduce clustering bias and reduce the impact of sequencing quality fluctuations on clustering results. Summary of the invention

[0005] In order to solve the shortcomings of the prior art, the embodiment of the present application provides an analysis system for the structure of tobacco fermentation microbial communities. The present application solves the technical problems that the prior art is prone to fall into local optimality when processing tobacco microbial sequences with complex hierarchical structures.

[0006] An embodiment of the present application provides an analysis system for the community structure of tobacco fermentation microorganisms, comprising: a sampling unit, a clustering unit and a community structure analysis unit; wherein the sampling unit is used to obtain microbial samples on the surface of tobacco leaves, so as to obtain original sequencing data through a high-throughput sequencing platform, and perform data preprocessing to obtain a sequence file; the clustering unit is used to process the preprocessed sequences using a distance-based hierarchical clustering method, and determine the target clustering results through a nonlinear distance update strategy and a clustering evaluation index; the community structure analysis unit is used to construct a phylogenetic classification tree based on the target clustering results, and obtain the fermentation microbial community structure characteristics of the tobacco sample in combination with the diversity index between samples.

[0007] In a possible implementation, microbial samples from the surface of tobacco leaves are obtained to obtain raw sequencing data through a high-throughput sequencing platform, and data preprocessing is performed, including: adjusting temperature and humidity parameters in a controlled environment, and using sterile tools to collect microbial samples from different areas on the surface of tobacco leaves; storing the collected samples in pre-cooled sterile sampling tubes; sequencing the stored samples using a high-throughput sequencing platform; transmitting and storing the sequences obtained by sequencing to form an initial data set; splicing the sequences in the paired initial data set based on sequence overlap, and classifying and organizing them according to sample identifiers after filtering to generate a standardized sequence file.

[0008] In a possible implementation, the preprocessed sequences are processed using a distance-based hierarchical clustering method, and the target clustering results are determined through a nonlinear distance update strategy and a clustering evaluation index, including: calculating the genetic distances between all sequences and forming a distance matrix; calculating the distance between units by treating each sequence as an independent clustering unit; dynamically adjusting the distance between clustering units using a nonlinear distance update strategy; selecting clustering units with the smallest distance to merge to form new clustering units; determining the optimal number of clusters using a clustering evaluation index; calculating the central sequence of each clustering unit; performing a consistency analysis on the determined central sequence to eliminate abnormal sequences; and classifying and summarizing the processed clustering results according to the source of the sample.

[0009] In a possible implementation, a phylogenetic classification tree is constructed based on the target clustering results, and the fermentation microbial community structure characteristics of the tobacco sample are obtained in combination with the diversity index between samples, including: selecting representative sequences from each operational classification unit for species classification annotation; constructing a phylogenetic classification tree by level-by-level classification, and calculating the relative abundance of each classification unit in different samples; calculating the diversity index between samples based on the abundance table, and sorting the samples using a multidimensional scaling analysis method; and constructing a clustering dendrogram by hierarchical clustering of the samples by identifying the dominant fungal groups at different processing stages.

[0010] In a possible implementation, the sequences in the paired initial data set are spliced ​​based on the sequence overlap relationship, and are classified and sorted according to the sample identifier after filtering to generate a standardized sequence file, including: splicing the paired sequences based on the sequence overlap relationship; filtering the sequence fragments below the quality threshold using a sliding window method; distinguishing samples based on the markers at both ends of the sequence and the primer sequence; correcting the sequence direction using a local alignment algorithm; and classifying and sorting the processed sequences according to the sample identifier to generate a standardized sequence file.

[0011] In a possible implementation, the genetic distances between all sequences are calculated and a distance matrix is ​​formed, including: obtaining a substitution distance value based on the product of the number of sequence base substitution sites and the substitution weight coefficient; obtaining a deletion distance value based on the product of the number of sequence base deletion sites and the deletion weight coefficient; obtaining an insertion distance value based on the product of the number of sequence base insertion sites and the insertion weight coefficient; dividing the sum of the substitution distance value, the deletion distance value and the insertion distance value by the maximum length value of the two sequences to obtain a sequence distance value; performing pairwise distance calculations on all sequences to generate a distance matrix.

[0012] In a possible implementation, a nonlinear distance update strategy is used to dynamically adjust the distance between clustering units, including: obtaining the distance value between two clustering units to be merged and other clustering units; taking the product of the smaller distance value of the two clustering units and the first update coefficient as the first update value; taking the product of the larger distance value of the two clustering units and the second update coefficient as the second update value; taking the product of the distance value between the two clustering units and the third update coefficient as the third update value; and taking the sum of the first update value, the second update value and the third update value as the updated distance value.

[0013] In a possible implementation, the optimal number of clusters is determined using a clustering evaluation index, including: calculating the average similarity of the internal sequence of each cluster unit; obtaining the minimum distance between each cluster unit and other units; calculating the ratio of the internal average similarity of each cluster unit to its minimum distance; dividing the sum of the ratios of all cluster units by the number of cluster units to obtain an evaluation score; when the evaluation score reaches a local maximum and the rate of change between two adjacent iterations is less than a preset threshold, it is determined to be the optimal number of clusters.

[0014] In the analysis system of a tobacco fermentation microorganism community structure provided as above, the embodiment of the present application determines the optimized weight coefficient configuration through the analysis of a large amount of tobacco fermentation microorganism sequence data, and improves the resolution of distance calculation. At the same time, by dividing the calculated distance value by the maximum length of the sequence involved in the comparison for standardization, the technical problem of incomparable distance values ​​between sequences of different lengths is solved. Further, in the sequence clustering process, the embodiment of the present application introduces a three-parameter dynamic weight model through a nonlinear distance update strategy, realizes the adaptive distance adjustment in the clustering process, can dynamically adjust the class distance according to the current clustering state, and effectively avoids the clustering deviation caused by the fixed update rule in the traditional method. Further, in some embodiments, the representative sequence is determined by minimizing the weighted average distance from the sequence to the center, wherein the weight coefficient is dynamically adjusted according to the sequence quality score, can effectively reduce the influence of sequencing quality fluctuation on the clustering result, and improve the reliability of the result. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A schematic diagram of an analysis system for tobacco fermentation microbial community structure provided in an embodiment of the present application; Figure 2 A schematic diagram of a sequence clustering processing method provided in an embodiment of the present application; Figure 3 A cluster comparison distribution diagram provided in an embodiment of the present application; Figure 4 A schematic diagram of the structural characteristics of a tobacco fermentation microbial community provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0018] Those skilled in the art can understand that the terms "first", "second" and the like in the embodiments of the present disclosure are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor represent the necessary logical order between them. It should also be understood that in the embodiments of the present disclosure, "multiple" can refer to two or more, and "at least one" can refer to one, two or more. It should also be understood that for any component, data or structure mentioned in the embodiments of the present disclosure, in the absence of explicit limitation or contrary revelation given in the context, it can generally be understood as one or more. In addition, the term "and / or" in the present disclosure is only a description of the association relationship of the associated objects, indicating that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are an "or" relationship. It should also be understood that the description of each embodiment in the present disclosure emphasizes the differences between the embodiments, and the same or similar parts can refer to each other. For the sake of brevity, they will not be repeated one by one.

[0019] At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present disclosure and its application or use. The techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices should be considered part of the specification. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] Figure 1 A schematic diagram of an analysis system for a tobacco fermentation microbial community structure provided in an embodiment of the present application. It should be understood that the system shown in the figure is exemplary and non-restrictive. This means that the system architecture involved is not limited to a specific form or design, but is presented as an example. In other words, the architecture shown in the figure can be regarded as a way of expression to clearly describe related concepts and relationships, and does not exclude other forms of architecture. Therefore, when explaining the architecture in the picture, it should be understood that the model has flexibility and diversity, and its purpose is to provide an exemplary description rather than a restrictive provision on a specific form.

[0022] It should be noted that the main technical problems faced by the cluster analysis of tobacco fermentation microbial community sequences include the high similarity of tobacco surface microbial sequences, large fluctuations in sequence quality, and complex data structures. These characteristics lead to large deviations in the traditional fixed parameter clustering method when processing such data. Specifically, when calculating sequence distances, since tobacco surface microbial sequences often show high similarity, it is difficult to accurately reflect the degree of difference between sequences using only a simple distance algorithm. The traditional edit distance algorithm uses the same weight coefficient for base substitution, deletion, and insertion, and this processing method cannot adapt to the characteristics of tobacco microbial sequences. The embodiment of the present application determines the optimized weight coefficient configuration through the analysis of a large amount of tobacco fermentation microbial sequence data. This differentiated weight setting can more accurately quantify the impact of sequence variation and improve the resolution of distance calculation. At the same time, by dividing the calculated distance value by the maximum length of the sequence involved in the comparison for standardization, the technical problem of incomparable distance values ​​between sequences of different lengths is solved.

[0023] In the process of sequence clustering, the complexity of tobacco fermentation microbial community data has revealed higher requirements for clustering algorithms. The traditional hierarchical clustering method adopts a linear distance update strategy, which is easy to fall into local optimality when processing tobacco microbial sequences with complex hierarchical structures. To solve this technical problem, the nonlinear distance update strategy of the embodiment of the present application introduces a three-parameter dynamic weight model to achieve adaptive distance adjustment in the clustering process. This nonlinear update mechanism can dynamically adjust the distance between classes according to the current clustering state, effectively avoiding the clustering bias caused by fixed update rules in traditional methods.

[0024] In view of the technical problem of fluctuation in the quality of microbial sequences during tobacco fermentation, the embodiment of the present application introduces a comprehensive evaluation mechanism based on sequence quality in the clustering result optimization stage. By calculating the ratio of the internal similarity of the clustering unit to the difference between the units, and normalizing it in combination with the number of clusters, an objective clustering quality evaluation system is established. When the evaluation index reaches the local maximum and the rate of change of adjacent iterations is less than 0.0173, the optimal number of clusters can be determined. This method solves the technical problem that the traditional fixed threshold method cannot adapt to sequence data of different qualities. In the central sequence selection link, the algorithm developed in the embodiment of the present application determines the representative sequence by minimizing the weighted average distance from the sequence to the center, and the weight coefficient is dynamically adjusted according to the sequence quality score. In this way, the influence of sequencing quality fluctuations on the clustering results is effectively reduced, and the reliability of the results is improved.

[0025] Specifically, the present application discloses a tobacco fermentation microbial community structure analysis system, including: a sampling unit 101, a clustering unit 102 and a community structure analysis unit 103.

[0026] The sampling unit 101 is used to obtain microbial samples on the surface of tobacco leaves, so as to obtain raw sequencing data through a high-throughput sequencing platform, and perform data preprocessing to obtain a sequence file. This includes: adjusting the temperature and humidity parameters in a controlled environment, and using sterile tools to collect microbial samples from different areas on the surface of tobacco leaves; storing the collected samples in pre-cooled sterile sampling tubes; sequencing the stored samples using a high-throughput sequencing platform; transmitting and storing the sequences obtained by sequencing to form an initial data set; splicing the sequences in the paired initial data set based on the sequence overlap relationship, and classifying and arranging them according to the sample identifier after filtering to generate a standardized sequence file.

[0027] In one implementation scenario, when collecting tobacco leaf samples and obtaining sequencing data, a controlled sampling environment must first be established. Specifically, the ambient temperature is stabilized at 23.5±0.3°C with the cooperation of the temperature control system, and the relative humidity is maintained within the range of 67.3±1.2% using a humidifying device. During sampling, a sterile sampling tool is used to collect microbial samples from the tobacco leaf surface in different regions, while ensuring that no exogenous contamination is introduced. Preferably, the sampling area includes the front and back of the tobacco leaf and the area near the veins to ensure the representativeness of the sample. The collected samples are immediately placed in a sterile sampling tube precooled at -20°C for storage. Subsequently, the stored samples are transferred to a high-throughput sequencing platform equipped with NovaSeq 6000 and other models for sequencing. During the sequencing process, the instrument operating temperature is maintained at 18-22°C, and the relative humidity is controlled in the range of 45-55% to ensure the accuracy of the sequencing data. In actual operation, the number of sequences obtained for each sample fluctuates between 60441-74962, and the sequence length is distributed in the range of 140-531bp. For example, in one embodiment, the raw data obtained by sequencing includes 834,184 valid sequences and 189,990,454 bases in total. These raw data are transmitted and stored in real time to form an initial data set.

[0028] Furthermore, the sequences in the paired initial data set are spliced ​​based on the sequence overlap relationship, and are classified and sorted according to the sample identifier after filtering to generate a standardized sequence file, including: splicing the paired sequences based on the sequence overlap relationship; filtering the sequence fragments below the quality threshold using a sliding window method; distinguishing samples based on the markers at both ends of the sequence and the primer sequence; correcting the sequence direction using a local alignment algorithm; and classifying and sorting the processed sequences according to the sample identifier to generate a standardized sequence file.

[0029] Specifically, when the original electronic data obtained is preprocessed, PE reads are first spliced ​​based on the Overlap relationship between sequences. During the specific implementation, the paired sequences in the original sequence are spliced ​​using the improved FLASH algorithm, and the minimum overlap length is set to 10bp during the splicing process, and the maximum mismatch rate does not exceed 10%. In order to ensure data quality, the Phred quality scoring system is introduced, and the quality threshold is set to Q20, that is, the sequencing error rate is less than 1%. For sequence fragments below the threshold, the sliding window method is used for filtering, and the window size is set to 4bp and the step length is 1bp. In the sequence optimization stage, samples are distinguished based on the Barcode markers and primer sequences specific to both ends of the sequence. Preferably, a double-end Barcode pairing verification strategy is used to ensure the accuracy of sample attribution. In the identified valid sequence, the sequence direction is determined by comparison analysis, and the Smith-Waterman local alignment algorithm is used to correct the sequence direction to obtain the optimized sequence. The processed sequence data is classified and sorted according to the sample-specific identifier to generate a standardized FASTQ format sequence file, which contains complete information such as sequence information, quality value, and sample identification.

[0030] The clustering unit 102 is used to process the preprocessed sequence using a distance-based hierarchical clustering method, and determine the target clustering result (such as Figure 2 ). These include: calculating the genetic distances between all sequences and forming a distance matrix; calculating the distances between units by treating each sequence as an independent clustering unit; dynamically adjusting the distances between clustering units by using a nonlinear distance update strategy; selecting clustering units with the smallest distance to merge into new clustering units; determining the optimal number of clusters by using clustering evaluation indicators; calculating the central sequence of each clustering unit; performing consistency analysis on the determined central sequence to eliminate abnormal sequences; and classifying and summarizing the processed clustering results by sample source.

[0031] The embodiment of the present application adopts a distance-based hierarchical clustering method with a sequence similarity of 97.3±0.2% as the clustering threshold. In specific implementation, the improved clustering algorithm in the UPARSE 7.0.1090 software package can be used, which is optimized in terms of memory usage and computing speed. The clustering process is divided into the following steps: first, the genetic distance between sequences is calculated, and the Kimura two-parameter model is used to calculate the genetic distance; then, based on the calculated distance matrix, UPGMA (unweighted pair average method) is used to construct a clustering tree; finally, the sequences are grouped according to the set similarity threshold to obtain operational taxonomic units (OTUs).

[0032] To improve clustering efficiency, the sequences can be pre-processed for bidirectional redundancy to remove identical sequences. The clustering results are stored in the form of a distance matrix file, which records the detailed genetic distance relationship between the sequences. At the same time, the abundance of each OTU is counted to generate an OTU abundance table containing the sample-OTU correspondence. The table is stored in a sparse matrix format and contains the specific sequence quantity information of each OTU in each sample.

[0033] The community structure analysis unit 103 is used to construct a phylogenetic classification tree based on the target clustering results, and to obtain the fermentation microbial community structure characteristics of the tobacco sample in combination with the diversity index between samples. This includes: selecting representative sequences from each operational classification unit to perform species classification annotation; constructing a phylogenetic classification tree by level-by-level sorting according to the classification, and calculating the relative abundance of each classification unit in different samples; calculating the diversity index between samples based on the abundance table, and sorting the samples using a multidimensional scaling analysis method; and constructing a clustering tree diagram by hierarchical clustering of the samples by identifying the dominant fungal groups at different processing stages.

[0034] Specifically, the most representative sequence was first selected from each OTU. The selection of representative sequences was determined based on the sequence quality score and the frequency of occurrence in the OTU. The selected representative sequences can be annotated for species classification using the RDP Classifier2.2 classifier software, which uses a Bayesian classification algorithm with a confidence threshold set to 0.8. The Unite Release 8.0 fungal database can be used as a reference database during the classification process, which contains expert-verified fungal classification information.

[0035] The annotation results were organized step by step according to the classification system of "kingdom, phylum, class, order, family, genus and species" to construct a phylogenetic classification tree. In the process of building the classification tree, the maximum likelihood method was used to estimate the evolutionary relationship between each taxonomic unit. Based on the annotation results, the relative abundance of each taxonomic unit in different samples was calculated to generate a multi-level species composition analysis table. At the same time, the diversity index of each group of samples was determined using an improved calculation method, including the Shannon index (reflecting community diversity), the Simpson index (reflecting dominance), the Ace index and the Chao index (reflecting species richness), as well as the Sobs index (reflecting the number of observed species) and the coverage index (reflecting sequencing depth).

[0036] Next, based on the multidimensional data obtained in the previous steps, the differences in community structure between different samples were analyzed. First, the Beta diversity index between samples was calculated based on the OTU abundance table, using the Bray-Curtis distance and UniFrac distance methods. The samples were then sorted using the non-metric multidimensional scaling (NMDS) method, and a two-dimensional sorting diagram was drawn to show the relationship between the samples. The NMDS analysis used a multiple random starting point strategy to select the sorting result with the smallest stress value.

[0037] In the community composition analysis, the dominant fungal groups at different processing stages were identified based on the species annotation results and abundance information. The relative abundance was calculated using the effective sequence normalization method, and species with an abundance of less than 0.01% were classified as "others". The samples were hierarchically clustered, and a UPGMA cluster dendrogram was constructed to display the phylogenetic relationship between the samples. Finally, LEfSe (linear discriminant analysis effect size) was used to systematically analyze the species with significant differences between samples at different processing stages, setting the LDA discrimination threshold to 2.0 and the significance level α to 0.05. Based on the analysis results, complete data containing community structure characteristics were formed.

[0038] Figure 2 A schematic diagram of a sequence clustering processing method provided in an embodiment of the present application.

[0039] like Figure 2 As shown, at step S201, the genetic distances between all sequences are calculated and a distance matrix is ​​formed. This includes: obtaining a replacement distance value based on the product of the number of sequence base replacement sites and the replacement weight coefficient; obtaining a deletion distance value based on the product of the number of sequence base deletion sites and the deletion weight coefficient; obtaining an insertion distance value based on the product of the number of sequence base insertion sites and the insertion weight coefficient; dividing the sum of the replacement distance value, the deletion distance value and the insertion distance value by the maximum length value of the two sequences to obtain a sequence distance value; performing pairwise distance calculations on all sequences to generate a distance matrix.

[0040] In one embodiment, the sequence distance calculation is first performed on the optimized sequence obtained in the second step. The optimized sequence data is imported into the calculation module, and the genetic distance between sequences is calculated using the improved Levenshtein distance algorithm, and the calculation formula is: D(i, j) = (L_sub×W_sub + L_del×W_del + L_ins×W_ins) / max(len_i, len_j), where D(i, j) represents the distance value between sequences i and j, L_sub, L_del, and L_ins represent the number of sites for base substitution, deletion, and insertion, respectively, W_sub=2.37, W_del=1.83, and W_ins=1.83 are the corresponding weight coefficients, and len_i and len_j are the lengths of the two sequences, respectively. The algorithm is used to calculate the distance between all sequences, forming an n×n distance matrix, where n is the total number of sequences.

[0041] At step S202, each sequence is used as an independent clustering unit to calculate the distance between the units. At step S203, a nonlinear distance update strategy is used to dynamically adjust the distance between the clustering units. This includes: obtaining the distance value between the two clustering units to be merged and other clustering units; taking the product of the smaller distance value of the two clustering units and the first update coefficient as the first update value; taking the product of the larger distance value of the two clustering units and the second update coefficient as the second update value; taking the product of the distance value between the two clustering units and the third update coefficient as the third update value; taking the sum of the first update value, the second update value and the third update value as the updated distance value.

[0042] Specifically, based on the calculated distance matrix, an improved hierarchical clustering method is used for sequence clustering. First, each sequence is regarded as an independent clustering unit, and the distances between all current clustering units are calculated. A nonlinear distance update strategy is used to dynamically adjust the distances between clustering units. The adjustment formula is: D_new(p,q) = α×min(D(p,r), D(q,r)) + β×max(D(p,r), D(q,r)) + γ×D(p,q), where D_new(p,q) is the updated distance value, p and q are the clustering units to be merged, r is the other clustering unit, and α=0.367, β=0.533, and γ=0.1 are distance update coefficients.

[0043] At step S204, the clustering unit with the smallest distance is selected for merging to form a new clustering unit. According to the updated distance value, two clustering units with the smallest distance are selected for merging to form a new clustering unit. The distance calculation, updating and merging process is repeated until all sequences are assigned to clustering units.

[0044] At step S205, the optimal number of clusters is determined using cluster evaluation indicators, which include: calculating the average similarity of the internal sequence of each cluster unit; obtaining the minimum distance between each cluster unit and other units; calculating the ratio of the internal average similarity of each cluster unit to its minimum distance; dividing the sum of the ratios of all cluster units by the number of cluster units to obtain an evaluation score; when the evaluation score reaches a local maximum and the change rate between two adjacent iterations is less than a preset threshold, it is determined as the optimal number of clusters.

[0045] Specifically, in the clustering process, an improved clustering evaluation index is introduced to determine the optimal number of clusters. The evaluation index comprehensively considers the internal similarity of the cluster unit and the difference between the units, and the calculation formula is: Score = (1 / k)×Σ(Si / Di), where k is the number of cluster units, Si is the average similarity of the sequences within the i-th cluster unit, and Di is the minimum distance between the i-th cluster unit and other units. When the Score value reaches the local maximum and the rate of change between two adjacent iterations is less than 0.0173, it is determined to be the optimal number of clusters. Based on the optimal number of clusters, all sequences are clustered to generate a clustering result file.

[0046] At step S206, the central sequence of each clustering unit is calculated. At step S207, the determined central sequence is subjected to consistency analysis to remove abnormal sequences. At step S208, the processed clustering results are classified and summarized according to the sample source.

[0047] In one embodiment, in order to post-process and optimize the clustering results. To calculate the central sequence of each clustering unit, an improved central sequence selection algorithm can be used to determine the representative sequence by minimizing the average distance from the sequence to the center. The calculation formula is: Center = argmin(Σ(D(s,c)×w(s))), where s is the sequence in the clustering unit, c is the candidate central sequence, and w(s) is the weight coefficient of the sequence, which is determined according to the sequence quality score. The determined central sequence is subjected to sequence consistency analysis, and abnormal sequences whose difference from the central sequence exceeds a threshold are eliminated. The processed clustering results are classified and summarized according to the sample source, and a result table containing information such as the clustering unit number, number of sequences, and representative sequence is generated.

[0048] Figure 3 A cluster comparison distribution diagram provided in the embodiment of the present application. It shows the multi-dimensional comparison results of a sequence clustering analysis method based on the improved Levenshtein distance algorithm and the traditional UPGMA clustering method. The analysis is based on experimental data of 834,184 valid sequences (total base number 189,990,454), and the sequence length is distributed in the range of 140-531bp.

[0049] During the sequence processing, the FLASH algorithm was used to splice PE reads, with the minimum overlap length set to 10bp and the maximum mismatch rate controlled within 15%. The scatter matrix shown in the figure reflects the distribution pattern of different clustering methods in the sequence feature space. The horizontal and vertical axes represent different characteristic parameters of the sequence, including key indicators such as sequence length, GC content, and quality distribution. The histogram on the diagonal shows the distribution characteristics of each parameter, and the scatter plot in the non-diagonal position shows the correlation between the parameters.

[0050] In the sequence distance calculation link, the improved Levenshtein algorithm introduced three key weight coefficients: base substitution weight W_sub=2.37, base deletion weight W_del=1.83, and base insertion weight W_ins=1.83. The selection of these weight coefficients makes the distance calculation results more accurately reflect the evolutionary relationship between sequences. It can be observed from the scatter plot that the improved algorithm (green and orange point groups) forms a tighter clustering structure in the feature space than the traditional UPGMA method (blue point group).

[0051] For the optimization of the clustering process, a nonlinear distance update strategy is adopted, which is dynamically adjusted through the combination of coefficients α=0.367, β=0.533, and γ=0.1. This update strategy not only maintains local similarity, but also takes into account the global community structure characteristics. As can be seen from the figure, the two main clusters (green and orange point groups) obtained by the improved algorithm have obvious boundaries in the feature space and a denser internal structure.

[0052] The clustering evaluation adopted a comprehensive indicator system, and the optimal number of clusters was determined when the change rate between two adjacent iterations was less than 0.0173. From different perspectives of the scatter matrix, it can be observed that the improved algorithm showed good clustering effects in all feature dimensions. In particular, in the relationship diagram between sequence length and GC content, the cluster boundaries obtained by the improved algorithm were clearer, indicating that this method can better identify the essential characteristics of the sequence.

[0053] The improved center sequence selection algorithm was used for post-processing of clustering results, and the representative sequence was determined by minimizing the average distance from the sequence to the center. From the distribution pattern of the scatter plot, it can be seen that this processing method effectively reduced the heterogeneity within the group. Combined with the RDP Classifier 2.2 classifier software (confidence threshold 0.8) and the annotation results of the Unite Release 8.0 fungal database, the dominant bacterial groups at different fermentation stages can be accurately identified.

[0054] Combined with experimental data analysis, this method showed good adaptability when processing sequences from different sources. The number of sequences obtained for each sample ranged from 60,441 to 74,962. Through improved distance calculation and clustering strategies, it can effectively reflect the differences in community structure between samples. The relative abundance calculation uses an effective sequence standardization method to classify species with an abundance of less than 0.01% into the "other" category. This processing method is more in line with the actual distribution characteristics of microbial communities.

[0055] It can be observed from the scatter matrix that the improved algorithm shows better clustering effects in multiple feature dimensions. In particular, in the feature space reflecting sequence similarity, the clusters obtained by the improved algorithm have higher cohesion and clearer boundaries. This result shows that by introducing differentiated weight coefficients and nonlinear distance update strategies, this method can more accurately characterize the structural characteristics of microbial communities.

[0056] based on Figure 3 According to the analysis results, the preferred implementation scheme of this application has achieved a good balance in the following key parameters: the minimum overlap length during sequence splicing is 10bp, the maximum mismatch rate is controlled within 15%; the distance calculation uses the weight coefficients of W_sub=2.37, W_del=1.83, and W_ins=1.83; the clustering process uses the nonlinear distance update coefficients of α=0.367, β=0.533, and γ=0.1; the iterative change rate threshold is set to 0.0173 in the clustering evaluation stage. This set of parameters can obtain biologically meaningful community structure analysis results while ensuring computational efficiency.

[0057] Figure 4 A schematic diagram of the structural characteristics of a tobacco fermentation microbial community provided in the examples of this application. Figure 4 As shown, this embodiment systematically analyzes the structural characteristics of the microbial community in the tobacco fermentation process based on high-throughput sequencing data. The sequencing process uses the NovaSeq 6000 platform under strictly controlled temperature (18-22°C) and relative humidity (45-55%) conditions. Through sequence processing, 60441-74962 high-quality sequences were obtained for each sample, with sequence lengths ranging from 140-531bp, a total of 834184 valid sequences, and a total number of bases of 189990454.

[0058] In the sequence clustering analysis, the improved FLASH algorithm was used to splice the original sequences, the minimum overlap length was set to 10 bp, and the maximum mismatch rate was controlled within 15%. The quality control was performed by the Phred quality scoring system (Q20 threshold), and the Smith-Waterman local alignment algorithm was used to correct the sequence direction. The sequence similarity of 97.3±0.2% was used as the clustering threshold, and the improved Levenshtein distance algorithm was used to calculate the genetic distance between sequences.

[0059] Figure 4 Two key features of community structure are demonstrated. Figure 4 (a) Results of principal coordinate analysis (PCoA), calculated based on Bray-Curtis distance. The first principal coordinate axis (PCoA1) explained 50.5% of the variance in community structure, and the second principal coordinate axis (PCoA2) explained 12.3% of the variance. From the scatter distribution, it can be observed that the S series samples (Q6-S to Q17-S) and the W series samples (Q1-W to Q17-W) formed a clear partition in the principal coordinate space. This distribution pattern reflects the significant effects of different treatments on the composition of the microbial community.

[0060] Figure 4 (b) The phylogenetic tree combined with the heat map shows the distribution characteristics of the main species in each sample. The clustering process adopts a nonlinear distance update strategy, introducing update coefficients of α=0.367, β=0.533, and γ=0.1, and dynamically adjusts the distance through the formula D_new(p,q) = α×min(D(p,r), D(q,r)) + β×max(D(p,r), D(q,r)) + γ×D(p,q). When the Score value reaches the local maximum and the adjacent iteration change rate is less than 0.0173, it is determined as the optimal number of clusters.

[0061] Data analysis showed that in the W series samples, Acinetobacter showed the highest relative abundance, accounting for an average of 32.8%. It was followed by hgcl_clade (23.7%) and Candidatus Aquiluna (14.2%). The species distribution in the S series samples was more even, mainly including Limnohabitans (8.3%), Luteolibacter (7.6%) and Thiobacillus (6.8%). This differential distribution reflects the selective growth characteristics of microbial communities under different treatment conditions.

[0062] Phylogenetic analysis was performed using RDP Classifier 2.2 classifier software for species annotation, with a confidence threshold of 0.8, and the Unite Release 8.0 fungal database was used as a reference. Characteristic species that contributed significantly to community differences were identified using the LEfSe analysis system (LDA threshold 2.0, α = 0.05). The results showed that the microbial communities in different treatment groups differed significantly in species composition and abundance distribution.

[0063] Beta diversity analysis used Bray-Curtis distance and UniFrac distance, combined with non-metric multidimensional scaling (NMDS) analysis to sort samples. The results showed that the Beta diversity index of samples between groups was significantly higher than that of samples within groups, further confirming the significant effect of treatment on microbial community structure. This difference in community structure provides an important basis for understanding the mechanism of action of microorganisms in the fermentation process.

[0064] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0065] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0066] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An analysis system for tobacco fermentation microbial community structure, characterized in that: include: Sampling unit, clustering unit and community structure analysis unit; among them, The sampling unit is used to obtain microbial samples on the surface of tobacco leaves, to obtain raw sequencing data through a high-throughput sequencing platform, and to perform data preprocessing to obtain a sequence file; The clustering unit is used to process the preprocessed sequence using a distance-based hierarchical clustering method, and determine the target clustering result through a nonlinear distance update strategy and a clustering evaluation index; The community structure analysis unit is used to construct a phylogenetic classification tree based on the target clustering result, and to obtain the fermentation microbial community structure characteristics of the tobacco sample in combination with the diversity index between samples.

2. The analysis system according to claim 1, characterized in that in, Obtain tobacco leaf surface microbial samples to obtain raw sequencing data through a high-throughput sequencing platform and perform data preprocessing, including: Temperature and humidity parameters were adjusted in a controlled environment, and sterile tools were used to collect microbial samples from different areas on the tobacco leaf surface; Store the collected samples in pre-cooled sterile sampling tubes; Sequencing the stored samples using a high-throughput sequencing platform; The sequences obtained by sequencing are transmitted and stored to form an initial data set; The sequences in the paired initial data set were spliced ​​based on sequence overlap and sorted according to sample identifiers after filtering to generate a standardized sequence file.

3. The analysis system according to claim 1, characterized in that in, The preprocessed sequences are processed using a distance-based hierarchical clustering method, and the target clustering results are determined through a nonlinear distance update strategy and clustering evaluation indicators, including: Calculate the genetic distances between all sequences and form a distance matrix; Each sequence was treated as an independent clustering unit to calculate the distance between units; A nonlinear distance updating strategy is used to dynamically adjust the distance between clustering units; Select the clustering unit with the smallest distance to merge into a new clustering unit; Determine the optimal number of clusters using clustering evaluation indicators; Calculate the center sequence of each cluster unit; Perform consistency analysis on the determined central sequence to eliminate abnormal sequences; The processed clustering results were classified and summarized according to the sample sources.

4. The analysis system according to claim 1, characterized in that in, A phylogenetic classification tree was constructed based on the target clustering results, and the fermentation microbial community structure characteristics of tobacco samples were obtained by combining the diversity index between samples, including: Select representative sequences from each operational taxonomic unit for species taxonomic annotation; The phylogenetic classification tree was constructed step by step according to the classification, and the relative abundance of each taxonomic unit in different samples was calculated; The diversity index among samples was calculated based on the abundance table, and the samples were ranked using multidimensional scaling analysis; The dominant fungal groups at different processing stages were identified and a cluster dendrogram was constructed by hierarchical clustering of the samples.

5. The analysis system according to claim 2, characterized in that in, Sequences in the paired initial datasets are spliced ​​based on sequence overlap and sorted by sample identifiers after filtering to generate standardized sequence files, including: The paired sequences are spliced ​​based on the sequence overlap relationship; The sliding window method was used to filter sequence fragments below the quality threshold; Samples are differentiated based on the markers at both ends of the sequence and the primer sequences; A local alignment algorithm was used to correct the sequence orientation; The processed sequences are classified and organized according to the sample identifiers to generate standardized sequence files.

6. The analysis system according to claim 3, characterized in that in, Calculate the genetic distances between all sequences and form a distance matrix, including: The substitution distance value is obtained based on the product of the number of sequence base substitution sites and the substitution weight coefficient; The missing distance value is obtained based on the product of the number of sequence base missing sites and the missing weight coefficient; The insertion distance value is obtained based on the product of the number of sequence base insertion sites and the insertion weight coefficient; Divide the sum of the replacement distance value, the deletion distance value and the insertion distance value by the maximum length value of the two sequences to obtain a sequence distance value; The pairwise distances of all sequences were calculated to generate a distance matrix.

7. The analysis system according to claim 3, characterized in that in, A nonlinear distance update strategy is used to dynamically adjust the distance between clustering units, including: Get the distance value between the two clustering units to be merged and other clustering units; Taking the product of the smaller distance value between the two clustering units and the first update coefficient as the first update value; taking the product of the larger distance value between the two clustering units and the second update coefficient as the second update value; taking the product of the distance value between the two clustering units and the third update coefficient as the third update value; The sum of the first updated value, the second updated value and the third updated value is used as the updated distance value.

8. The analysis system according to claim 3, characterized in that in, The optimal number of clusters is determined using clustering evaluation indicators, including: Calculate the average similarity of the sequences within each clustering unit; Get the minimum distance between each cluster unit and other units; Calculate the ratio of the internal average similarity of each clustering unit to its minimum distance; The sum of the ratios of all cluster units is divided by the number of cluster units to obtain the evaluation score; When the evaluation score reaches a local maximum and the change rate between two adjacent iterations is less than a preset threshold, it is determined as the optimal cluster number.

Images